INCORPORATING BERT INTO NEURAL MACHINE TRANSLATION - OpenReview

 
CONTINUE READING
Published as a conference paper at ICLR 2020

I NCORPORATING BERT INTO
N EURAL M ACHINE T RANSLATION
 Jinhua Zhu1,∗ , Yingce Xia2,∗, Lijun Wu3 , Di He4 ,
 Tao Qin2 , Wengang Zhou1 , Houqiang Li1 , Tie-Yan Liu2
 1
 CAS Key Laboratory of GIPAS, EEIS Department, University of Science and Technology of China;
 2
 Microsoft Research;
 3
 Sun Yat-sen University;
 4
 Key Laboratory of Machine Perception (MOE), School of EECS, Peking University
 1
 teslazhu@mail.ustc.edu.cn, {zhwg,lihq}@ustc.edu.cn
 2
 yingce.xia@gmail.com, {taoqin,tyliu}@microsoft.com
 3
 wulijun3@mail2.sysu.edu.cn 4 di he@pku.edu.cn

 A BSTRACT
 The recently proposed BERT (Devlin et al., 2019) has shown great power on a va-
 riety of natural language understanding tasks, such as text classification, reading
 comprehension, etc. However, how to effectively apply BERT to neural machine
 translation (NMT) lacks enough exploration. While BERT is more commonly
 used as fine-tuning instead of contextual embedding for downstream language
 understanding tasks, in NMT, our preliminary exploration of using BERT as con-
 textual embedding is better than using for fine-tuning. This motivates us to think
 how to better leverage BERT for NMT along this direction. We propose a new
 algorithm named BERT-fused model, in which we first use BERT to extract rep-
 resentations for an input sequence, and then the representations are fused with
 each layer of the encoder and decoder of the NMT model through attention mech-
 anisms. We conduct experiments on supervised (including sentence-level and
 document-level translations), semi-supervised and unsupervised machine trans-
 lation, and achieve state-of-the-art results on seven benchmark datasets. Our code
 is available at https://github.com/bert-nmt/bert-nmt.

1 I NTRODUCTION
Recently, pre-training techniques, like ELMo (Peters et al., 2018), GPT/GPT-2 (Radford et al., 2018;
2019), BERT (Devlin et al., 2019), cross-lingual language model (briefly, XLM) (Lample & Con-
neau, 2019), XLNet (Yang et al., 2019b) and RoBERTa (Liu et al., 2019) have attracted more and
more attention in machine learning and natural language processing communities. The models are
first pre-trained on large amount of unlabeled data to capture rich representations of the input, and
then applied to the downstream tasks by either providing context-aware embeddings of an input se-
quence (Peters et al., 2018), or initializing the parameters of the downstream model (Devlin et al.,
2019) for fine-tuning. Such pre-training approaches lead to significant improvements on natural lan-
guage understanding tasks. Among them, BERT is one of the most powerful techniques that inspires
lots of variants like XLNet, XLM, RoBERTa and achieves state-of-the-art results for many language
understanding tasks including reading comprehension, text classification, etc (Devlin et al., 2019).
Neural Machine Translation (NMT) aims to translate an input sequence from a source language to a
target language. An NMT model usually consists of an encoder to map an input sequence to hidden
representations, and a decoder to decode hidden representations to generate a sentence in the target
language. Given that BERT has achieved great success in language understanding tasks, a question
worthy studying is how to incorporate BERT to improve NMT. Due to the computation resource
limitation, training a BERT model from scratch is unaffordable for many researchers. Thus, we
focus on the setting of leveraging a pre-trained BERT model (instead of training a BERT model
from scratch) for NMT.
 ∗
 This work is conducted at Microsoft Research Asia. The first two authors contributed equally to this work.

 1
Published as a conference paper at ICLR 2020

Given that there is limited work leveraging BERT for NMT, our first attempt is to try two previous
strategies: (1) using BERT to initialize downstream models and then fine-tuning the models, and (2)
using BERT as context-aware embeddings for downstream models. For the first strategy, follow-
ing Devlin et al. (2019), we initialize the encoder of an NMT model with a pre-trained BERT model,
and then finetune the NMT model on the downstream datasets. Unfortunately, we did not observe
significant improvement. Using a pre-trained XLM (Lample & Conneau, 2019) model, a variant of
BERT for machine translation, to warm up an NMT model is another choice. XLM has been ver-
ified to be helpful for WMT’16 Romanian-to-English translation. But when applied to a language
domain beyond the corpus for training XLM (such as IWSLT dataset (Cettolo et al., 2014), which
is about spoken languages) or when large bilingual data is available for downstream tasks, no sig-
nificant improvement is observed neither. For the second strategy, following the practice of (Peters
et al., 2018), we use BERT to provide context-aware embeddings for the NMT model. We find that
this strategy outperforms the first one (please refer to Section 3 for more details). This motivates us
to go along this direction and design more effective algorithms.
We propose a new algorithm, BERT-fused model, in which we exploit the representation from BERT
by feeding it into all layers rather than served as input embeddings only. We use the attention
mechanism to adaptively control how each layer interacts with the representations, and deal with the
case that BERT module and NMT module might use different word segmentation rules, resulting in
different sequence (i.e., representation) lengths. Compared to standard NMT, in addition to BERT,
there are two extra attention modules, the BERT-encoder attention and BERT-decoder attention. An
input sequence is first transformed into representations processed by BERT. Then, by the BERT-
encoder attention module, each NMT encoder layer interacts with the representations obtained from
BERT and eventually outputs fused representations leveraging both BERT and the NMT encoder.
The decoder works similarly and fuses BERT representations and NMT encoder representations.
We conduct 14 experiments on various NMT tasks to verify our approach, including supervised,
semi-supervised and unsupervised settings. For supervised NMT, we work on five tasks of IWSLT
datasets and two WMT datasets. Specifically, we achieve 36.11 BLEU score on IWSLT’14 German-
to-English translation, setting a new record on this task. We also work on two document-level
translations of IWSLT, and further boost the BLEU score of German-to-English translation to 36.69.
On WMT’14 datasets, we achieve 30.75 BLEU score on English-to-German translation and 43.78 on
English-to-French translation, significantly better over the baselines. For semi-supervised NMT, we
boost BLEU scores of WMT’16 Romanian-to-English translation with back translation (Sennrich
et al., 2016b), a classic semi-supervised algorithm, from 37.73 to 39.10, achieving the best result
on this task. Finally, we verify our algorithm on unsupervised English↔French and unsupervised
English↔Romanian translations and also achieve state-of-the-art results.

2 BACKGROUND AND RELATED WORK

We briefly introduce the background of NMT and review current pre-training techniques.
NMT aims to translate an input sentence from the source language to the target one. An NMT
model usually consists of an encoder, a decoder and an attention module. The encoder maps the
input sequence to hidden representations and the decoder maps the hidden representations to the
target sequence. The attention module is first introduced by Bahdanau et al. (2015), which is used
to better align source words and target words. The encoder and decoder can be specialized as
LSTM (Hochreiter & Schmidhuber, 1997; Sutskever et al., 2014; Wu et al., 2016), CNN (Gehring
et al., 2017) and Transformer (Vaswani et al., 2017). A Transformer layer consists of three sub-
layers, a self-attention layer that processes sequential data taking the context of each timestep into
consideration, an optional encoder-decoder attention layer that bridges the input sequence and tar-
get sequence which exists in decoder only, and a feed-forward layer for non-linear transformation.
Transformer achieves the state-of-the-art results for NMT (Barrault et al., 2019). In this work, we
will use Transformer as the basic architecture of our model.
Pre-training has a long history in machine learning and natural language processing (Erhan et al.,
2009; 2010). Mikolov et al. (2013) and Pennington et al. (2014) proposed to use distributional
representations (i.e., word embeddings) for individual words. Dai & Le (2015) proposed to train
a language model or an auto-encoder with unlabeled data and then leveraged the obtained model
to finetune downstream tasks. Pre-training has attracted more and more attention in recent years

 2
Published as a conference paper at ICLR 2020

and achieved great improvements when the data scale becomes large and deep neural networks
are employed. ELMo was proposed in Peters et al. (2018) based on bidirectional LSTMs and its
pre-trained models are fed into downstream tasks as context-aware inputs. In GPT (Radford et al.,
2018), a Transformer based language model is pre-trained on unlabeled dataset and then finetuned
on downstream tasks. BERT (Devlin et al., 2019) is one of the widely adopted pre-training approach
for model initialization. The architecture of BERT is the encoder of Transformer (Vaswani et al.,
2017). Two kinds of objective functions are used in BERT training: (1) Masked language modeling
(MLM), where 15% words in a sentence are masked and BERT is trained to predict them with their
surrounding words. (2) Next sentence prediction (NSP): Another task of pre-training BERT is to
predict whether two input sequences are adjacent. For this purpose, the training corpus consists
of tuples ([cls], input 1, [sep], input 2, [sep]), with learnable special tokens [cls]
to classify whether input 1 and input 2 are adjacent and [sep] to segment two sentences, and
with probability 50%, the second input is replaced with a random input. Variants of BERT have been
proposed: In XLM (Lample & Conneau, 2019), the model is pre-trained based on multiple languages
and NSP task is removed; in RoBERTa (Liu et al., 2019), more unlabeled data is leveraged without
NSP task neither; in XLNet (Yang et al., 2019b), a permutation based modeling is introduced.

3 A PRELIMINARY EXPLORATION

While a few pieces of work (Lample & Conneau, 2019; Song et al., 2019) design specific pre-
training methods for NMT, they are time and resource consuming given that they need to pre-train
large models from scratch using large-scale data, and even one model for each language pair. In this
work, we focus on the setting of using a pre-trained BERT model. Detailed model download links
can be found in Appendix D.
Considering that pre-trained models have been utilized in two different ways for other natural lan-
guage tasks, it is straightforward to try them for NMT. Following previous practice, we make the
following attempts.
(I) Use pre-trained models to initialize the NMT model. There are different implementations for this
approach. (1) Following (Devlin et al., 2019), we initialize the encoder of an NMT model with a pre-
trained BERT. (2) Following (Lample & Conneau, 2019), we initialize the encoder and/or decoder
of an NMT model with XLM.
(II) Use pre-trained models as inputs to the NMT model. Inspired from (Peters et al., 2018), we feed
the outputs of the last layer of BERT to an NMT model as its inputs.
We conduct experiments on the IWSLT’14 English→German translation, a widely adopted dataset
for machine translation consisting of 160k labeled sentence pairs. We choose Transformer (Vaswani
et al., 2017) as the basic model architecture with transformer iwslt de en configuration (a
six-layer model with 36.7M parameters). The translation quality is evaluated by BLEU (Papineni
et al., 2002) score; the larger, the better. Both BERTbase and XLM models are pre-trained and we
get them from the Web. More details about the experimental settings are included in Appendix A.2.

 Table 1: Preliminary explorations on IWSLT’14 English→German translation.
 Algorithm BLEU score
 Standard Transformer 28.57
 Use BERT to initialize the encoder of NMT 27.14
 Use XLM to initialize the encoder of NMT 28.22
 Use XLM to initialize the decoder of NMT 26.13
 Use XLM to initialize both the encoder and decoder of NMT 28.99
 Leveraging the output of BERT as embeddings 29.67

The results are shown in Table 1. We have several observations: (1) Using BERT to initialize the en-
coder of NMT can only achieve 27.14 BLEU score, which is even worse than standard Transformer
without using BERT. That is, simply using BERT to warm up an NMT model is not a good choice.
(2) Using XLM to initialize the encoder or decoder respectively, we get 28.22 or 26.13 BLEU score,
which does not outperform the baseline. If both modules are initialized with XLM, the BLEU score

 3
Published as a conference paper at ICLR 2020

is boosted to 28.99, slightly outperforming the baseline. Although XLM achieved great success
on WMT’16 Romanian-to-English, we get limited improvement here. Our conjecture is that the
XLM model is pre-trained on news data, which is out-of-domain for IWSLT dataset mainly about
spoken languages and thus, leading to limited improvement. (3) When using the output of BERT
as context-aware embeddings of the encoder, we achieve 29.67 BLEU, much better than using pre-
trained models for initialization. This shows that leveraging BERT as a feature provider is more
effective in NMT. This motivates us to take one step further and study how to fully exploit such
features provided by pre-trained BERT models.

4 A LGORITHM

In this section, we first define the necessary notations, then introduce our proposed BERT-fused
model and finally provide discussions with existing works.
Notations Let X and Y denote the source language domain and target language domain respectively,
which are the collections of sentences with the corresponding languages. For any sentence x ∈ X
and y ∈ Y, let lx and ly denote the number of units (e.g., words or sub-words) in x and y. The
i-th unit in x/y is denoted as xi /yi . Denote the encoder, decoder and BERT as Enc, Dec and BERT
respectively. For ease of reference, we call the encoder and decoder in our work as the NMT module.
W.l.o.g., we assume both the encoder and decoder consists of L layers. Let attn(q, K, V ) denote
the attention layer, where q, K and V indicate query, key and value respectively (Vaswani et al.,
2017). We use the same feed-forward layer as that used in (Vaswani et al., 2017) and denote it as
FFN. Mathematical formulations of the above layers are left at Appendix E.

4.1 BERT- FUSED MODEL

An illustration of our algorithm is shown in Figure 1. Any input x ∈ X is progressively processed
by the BERT, encoder and decoder.

 decoder
 × Add & Norm
 layer
 Feed
 
 Forward

 × encoder Add & Norm
 layer Add & Norm

 Feed BERT-Dec Enc-Dec
 Forward Attention Attention
 
 Add & Norm
 …… Add & Norm
 BERT-Enc Self
 Attention Attention Self
 Attention
 BERT
 −1 −1

Figure 1: The architecture of BERT-fused model. The left and right figures represent the BERT,
 L
encoder and decoder respectively. Dash lines denote residual connections. HB (red part) and HE
(green part) denote the output of the last layer from BERT and encoder.

Step-1: Given any input x ∈ X , BERT first encodes it into representation HB = BERT(x). HB is
the output of the last layer in BERT. The hB,i ∈ HB is the representation of the i-th wordpiece in x.
 l 0
Step-2: Let HE denote the hidden representation of l-th layer in the encoder, and let HE denote
 l l
word embedding of sequence x. Denote the i-th element in HE as hi for any i ∈ [lx ]. In the l-th

 4
Published as a conference paper at ICLR 2020

layer, l ∈ [L],
 1
 h̃li = attnS (hl−1 l−1 l−1 l−1
 
 i , HE , HE ) + attnB (hi , HB , HB ) , ∀i ∈ [lx ], (1)
 2
where attnS and attnB are attention models (see Eqn.(6)) with different parameters. Then each
 l
h̃li is further processed by FFN(·) defined in Eqn.(7) and we get the output of the l-th layer: HE =
 l l L
(FFN(h̃1 ), · · · , FFN(h̃lx )). The encoder will eventually output HE from the last layer.
 l
Step-3: Let S
Published as a conference paper at ICLR 2020

Limitations We are aware that our approach has several limitations. (1) Additional storage cost: our
approach leverages a BERT model, which results in additional storage cost. However, considering
the BLEU improvement and the fact that we do not need additional training of BERT, we believe
that the additional storage is acceptable. (2) Additional inference time: We use BERT to encode the
input sequence, which takes about 45% additional time (see Appendix C for details). We will leave
the improvement of the above two limitations as future work.

5 A PPLICATION TO SUPERVISED NMT AND SEMI - SUPERVISED NMT

We first verify our BERT-fused model on the supervised setting, including low-resource and rich-
resource scenarios. Then we conduct experiments on document-level translation to verify our ap-
proach. Finally, we combine BERT-fused model with back translation (Sennrich et al., 2016b) to
verify the effectiveness of our method on semi-supervised NMT.

5.1 S ETTINGS

Dataset For the low-resource scenario, we choose IWSLT’14 English↔German (En↔De),
English→Spanish (En→Es), IWSLT’17 English→French (En→Fr) and English→Chinese
(En→Zh) translation. There are 160k, 183k, 236k, 235k bilingual sentence pairs for En↔De,
En→Es, En→Fr and En→Zh tasks. Following the common practice (Edunov et al., 2018), for
En↔De, we lowercase all words. All sentences are preprocessed by BPE (Sennrich et al., 2016c).
The model configuration is transformer iwslt de en, representing a six-layer model with
embedding size 512 and FFN layer dimension 1024. For the rich-resource scenario, we work on
WMT’14 En→De and En→Fr, whose corpus sizes are 4.5M and 36M respectively. We concate-
nate newstest2012 and newstest2013 as the validation set and use newstest2014 as the test set. The
model configuration is transformer big, another six-layer network with embedding size 1024
and FFN layer dimension 4096. More details about data and model are left in Appendix A.1.
We choose BERTbase for IWSLT tasks and BERTlarge for WMT tasks, which can ensure that the
dimension of the BERT and NMT model almost match. The BERT models are fixed during training.
Detailed BERT information for each task is in Appendix D. The drop-net rate pnet is set as 1.0.
Training Strategy We first train an NMT model until convergence, then initialize the encoder and
decoder of the BERT-fused model with the obtained model. The BERT-encoder attention and BERT-
decoder attention are randomly initialized. Experiments on IWSLT and WMT tasks are conducted
on 1 and 8 M40 GPUs respectively. The batchsize is 4k tokens per GPU. Following (Ott et al., 2018),
for WMT tasks, we accumulate the gradient for 16 iterations and then update to simulate a 128-GPU
environment. It takes 1, 8 and 14 days to obtain the pre-trained NMT models, and additional 1, 7
and 10 days to finish the whole training process. The optimization algorithm is Adam (Kingma &
Ba, 2014) with initial learning rate 0.0005 and inverse sqrt learning rate scheduler (Vaswani
et al., 2017). For WMT’14 En→De, we use beam search with width 4 and length penalty 0.6 for
inference following (Vaswani et al., 2017). For other tasks, we use width 5 and length penalty 1.0.
Evaluation We use multi-bleu.perl to evaluate IWSLT’14 En↔De and WMT translation
tasks for fair comparison with previous work. For the remaining tasks, we use a more advance
implementation of BLEU score, sacreBLEU for evaluation. Script urls are in Appendix A.1.

5.2 R ESULTS

The results of IWSLT translation tasks are reported in Ta-
 Table 2: BLEU of all IWSLT tasks.
ble 2. We implemented standard Transformer as baseline.
 Transformer BERT-fused
Our proposed BERT-fused model can improve the BLEU
scores of the five tasks by 1.88, 1.47, 2.4, 1.9 and 2.8 En→De 28.57 30.45
points respectively, demonstrating the effectiveness of our De→En 34.64 36.11
method. The consistent improvements on various tasks En→Es 39.0 41.4
shows that our method works well for low-resource trans- En→Zh 26.3 28.2
 En→Fr 35.9 38.7
lations. We achieved state-of-the-art results on IWSLT’14
De→En translation, a widely investigated baseline in ma-

 6
Published as a conference paper at ICLR 2020

chine translation. The comparison with previous methods are shown in Appendix B.4 due to space
limitation.
The results of WMT’14 En→De and En→Fr are shown in Table 3. Our reproduced Transformer
matches the results reported in Ott et al. (2018), and we can see that our BERT-fused model can
improve these two numbers to 30.75 and 43.78, achieving 1.63 and 0.82 points improvement. Our
approach also outperforms the well-designed model DynamicConv (Wu et al., 2019) and a model
obtained through neural architecture search (So et al., 2019).

 Table 3: BLEU scores of WMT’14 translation.
 Algorithm En→De En→Fr
 DynamicConv (Wu et al., 2019) 29.7 43.2
 Evolved Transformer (So et al., 2019) 29.8 41.3
 Transformer + Large Batch (Ott et al., 2018) 29.3 43.0
 Our Reproduced Transformer 29.12 42.96
 Our BERT-fused model 30.75 43.78

5.3 T RANSLATION WITH DOCUMENT- LEVEL CONTEXTUAL INFORMATION

BERT is able to capture the relation between two sentences, since the next sentence prediction (NSP)
task is to predict whether two sentences are adjacent. We can leverage this property to improve
translation with document-level contextual information (Miculicich et al., 2018), which is briefly
denoted as document-level translation. The inputs are a couple of sentences extracted from a para-
graph/document, xd1 , xd2 , · · · , xdT , where the T x’s are contextually correlated. We want to translate
them into target language by considering the contextual information.
Algorithm In our implementation, to translate a sentence x to target domain, we leverage the con-
textual information by taking both x and its preceding sentence xprev as inputs. x is fed into Enc,
which is the same as sentence-level translation. For the input of BERT, it is the concatenation of two
sequences: ([cls], xprev , [sep], x, [sep]), where both [cls] and [sep] are special tokens
of BERT.
Setting We use IWSLT’14 En↔De dataset as introduced in Section 5.1. The data is a collection
of TED talks, where each talk consists of several sequences. We can extract the adjacent sentences
for training, validation and test sets. The training strategy, hyperparameter selection and evaluation
metric are the same for sentence-level translation.
Baselines We use two baselines here. (1) To Table 4: BLEU of document-level translation.
demonstrate how BERT works in our model, we En→De De→En
replace BERT by a Transformer with configu- Sentence-level 28.57 34.64
ration transformer iwslt de en, which
is randomly initialized and jointly trained. (2) Our Document-level 28.90 34.95
Another baseline is proposed by Miculicich Miculicich et al. (2018) 27.94 33.97
et al. (2018), where multiple preceding sen- Sentence-level + BERT 30.45 36.11
tences in a document are leveraged using a hi- Document-level + BERT 31.02 36.69
erarchical attention network.
Results The results are shown in Table 4. We can see that introducing contextual information from
an additional encoder can boost the sentence-level baselines, but the improvement is limited (0.33
for En→De and 0.31 for De→En). For Miculicich et al. (2018), the best results we obtain are 27.94
and 33.97 respectively, which are worse than the sentence-level baselines. Combining BERT-fused
model and document-level information, we can eventually achieve 31.02 for En→De and 36.69 for
De→En. We perform significant test1 between sentence-level and document-level translation. Our
document-level BERT-fused model significantly outperforms sentence-level baseline with p-value
less than 0.01. This shows that our approach not only works for sentence-level translation, but can
also be generalized to document-level translation.
 1
 https://github.com/moses-smt/mosesdecoder/blob/master/scripts/
analysis/bootstrap-hypothesis-difference-significance.pl

 7
Published as a conference paper at ICLR 2020

5.4 A PPLICATION TO SEMI - SUPERVISED NMT

We work on WMT’16 Romanian→English (Ro→En) translation to verify whether our approach can
still make improvement over back translation (Sennrich et al., 2016b), the standard and powerful
semi-supervised way to leverage monolingual data in NMT.
The number of bilingual sentence pairs for Ro→En is 0.6M . Sennrich et al. (2016a) provided
2M back translated data2 . We use newsdev2016 as validation set and newstest2016 as test set.
Sentences were encoded using BPE with a shared source-target vocabulary of about 32k tokens. We
use transformer big configuration. Considering there is no Romanian BERT, we use the cased
multilingual BERT (please refer to Appendix D) to encode inputs. The drop-net rate pnet is set as
1.0. The translation quality is evaluated by multi-bleu.perl.
The results are shown in Table 5. The Table 5: BLEU scores of WMT’16 Ro→En.
Transformer baseline achieves 33.12 BLEU Methods BLEU
score. With back-translation, the performance
is boosted to 37.73. We use the model obtained Sennrich et al. (2016a) 33.9
with back-translation to initialize BERT-fused XLM (Lample & Conneau, 2019) 38.5
model, and eventually reach 39.10 BLEU. Such Standard Transformer 33.12
a score surpasses the previous best result 38.5 + back translation 37.73
achieved by XLM (Lample & Conneau, 2019) + BERT-fused model 39.10
and sets a new record. This demonstrates that
our proposed approach is effective and can still achieve improvement over strong baselines.

6 A BLATION STUDY

We conduct two groups of ablation studies on IWSLT’14 En→De translation to better understand
our model.

 Table 6: Ablation study on IWSLT’14 En→De.
 Standard Transformer 28.57
 BERT-fused model 30.45
 Randomly initialize encoder/decoder of BERT-fused model 27.03
 Jointly tune BERT and encoder/decoder of BERT-fused model 28.87
 Feed BERT feature into all layers without attention 29.61
 Replace BERT output with random vectors 28.91
 Replace BERT with the encoder of another Transformer model 28.99
 Remove BERT-encoder attention 29.87
 Remove BERT-decoder attention 29.90

Study for training strategy and network architecture
We conduct ablation study to investigate the performance of each component of our model and
training strategy. Results are reported in Table 6:
(1) We randomly initialize the NMT module (i.e., encoder and decoder) of BERT-fused model in-
stead of using a warm-start one as introduced in the training strategy of Section 5.1. In this way, we
can only achieve 27.03 BLEU score, which cannot catch up with the baseline. We also jointly train
BERT model with the NMT module. Although it can also boost the baseline from 28.57 to 28.87, it
is not as good as fixing the BERT part, whose BLEU is 30.45.
(2) We feed the output of BERT into all layers of the encoder without attention models. That is,
the Eqn.(1) is revised to h̃li = 21 attnS (hl−1 l−1 l−1 l l−1
 
 i , HE , HE ) + W h
 B i ) , where WBl is learnable.
In this case, the encoder and BERT have to share the same vocabulary. The BLEU score is 29.61,
which is better than the standard Transformer but slightly worse than leveraging the output of BERT
 2
 Data at http://data.statmt.org/rsennrich/wmt16_backtranslations/ro-en/.

 8
Published as a conference paper at ICLR 2020

as embedding. This shows that the output of BERT should not be fused into each layer directly, and
using the attention model to bridge the relation is better than using simple transformation. More
results on different languages are included in Appendix B.3. To illustrate the effectiveness of our
method, we choose another two kinds of ways to encode the input sequence rather than using BERT:
(1) Using a fixed and randomly initialized embedding; (2) Using the encoder from another NMT
model. Their BLEU scores are 28.91 and 28.99 respectively, indicating that the BERT pre-trained
on large amount of unlabeled data can provide more helpful features to NMT.
(3) To verify where the output of BERT should be connected to, we remove the BERT-encoder atten-
tion (i.e., attnB in Eqn.(1)) and the BERT-decoder attention (i.e,, attnB in Eqn.(2)) respectively.
Correspondingly, the BLEU score drops from 30.45 to 29.87 and 29.90. This indicates that the out-
put of BERT should be leveraged by both encoder and decoder to achieve better performances. At
last, considering that there are two stacked encoders in our model, we also choose ensemble models
and deeper NMT models as baselines. Our approach outperforms the above baselines. The results
are left in Appendix B.2 due to space limitation.
Study on drop-net
To investigate the effect of drop-net, we conduct experiments on IWSLT’14 En→De dataset with
different drop-net probability, pnet ∈ {0, 0.2, 0.4, 0.6, 0.8, 1.0}. The results are shown in Figure 2.
As can been seen, although larger pnet leads to larger training loss, it leads to smaller validation loss
and so better BLUE scores. This shows that the drop-net trick can indeed improve the generalization
ability of our model. We fix pnet = 1.0 in other experiments unless specially specified.

 6 4.5 32
 pnet 0.0 pnet 0.6 pnet 0.0 pnet 0.6
 pnet 0.2 pnet 0.8 pnet 0.2 pnet 0.8
 pnet 0.4 pnet 1.0 pnet 0.4 pnet 1.0
LOSS

 LOSS

 4 4.0 BLEU
 30

 pnet 0.0 pnet 0.6
 pnet 0.2 pnet 0.8
 pnet 0.4 pnet 1.0
 20 20 40 60 80 100 120 3.50 20 40 60 80 100 120 280 20 40 60 80 100 120
 EPOCH EPOCH EPOCH
 (a) Training loss. (b) Validation loss. (c) Validation BLEU.

 Figure 2: Training/validation curves with different pnet ’s.

7 A PPLICATION TO UNSUPERVISED NMT

We work on unsupervised En↔Fr and En↔Ro translation. The data processing, architecture selec-
tion and training strategy is the same as Lample & Conneau (2019).
Settings For En↔Fr, we use 190M monolingual English sentences and 62M monolingual French
sentences from WMT News Crawl datasets, which is the same as that used in (Song et al., 2019).3
For unsupervised En↔Ro translation, we use 50M English sentences from News Crawl (sampled
from the data for En→Fr) and collect 2.9M sentences for Romanian by concatenating News Crawl
data sets and WMT’16 Romanian monolingual data following Lample et al. (2018). The data is
preprocessed in the same way as Lample & Conneau (2019).
We use the same model configuration as Lample & Conneau (2019), with details in Appendix A.3.
The BERT is the pre-trained XLM model (see Appendix D). We first train an unsupervised NMT
model following Lample & Conneau (2019) until convergence. Then we initialize our BERT-fused
model with the obtained model and continue training. We train models on 8 M40 GPUs, and the
batchsize is 2000 tokens per GPU. We use the same optimization hyper-parameters as that described
in Lample & Conneau (2019).

 3
 Data source: https://modelrelease.blob.core.windows.net/mass/en-fr.tar.gz.

 9
Published as a conference paper at ICLR 2020

 Table 7: BLEU scores of unsupervised NMT.
 En→Fr Fr→En En→Ro Ro→En
 Lample et al. (2018) 27.6 27.7 25.1 23.9
 XLM (Lample & Conneau, 2019) 33.4 33.3 33.3 31.8
 MASS (Song et al., 2019) 37.50 34.90 35.20 33.10
 Our BERT-fused model 38.27 35.62 36.02 33.20

Results The results of unsupervised NMT are shown in Table 7. With our proposed BERT-fused
model, we can achieve 38.27, 35.62, 36.02 and 33.20 BLEU scores on the four tasks, setting state-
of-the-art results on these tasks. Therefore, our BERT-fused model also benefits unsupervised NMT.

8 C ONCLUSION AND FUTURE WORK

In this work, we propose an effective approach, BERT-fused model, to combine BERT and NMT,
where the BERT is leveraged by the encoder and decoder through attention models. Experiments on
supervised NMT (including sentence-level and document-level translations), semi-supervised NMT
and unsupervised NMT demonstrate the effectiveness of our method.
For future work, there are many interesting directions. First, we will study how to speed up in-
ference time. Second, we can apply such an algorithm to more applications, like questioning and
answering. Third, how to compress BERT-fused model into a light version is another topic. There
are some contemporary works leveraging knowledge distillation to combine pre-trained models with
NMT (Yang et al., 2019a; Chen et al., 2019), which is a direction to explore.

R EFERENCES
Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translation by jointly
 learning to align and translate. In 6th International Conference on Learning Representations,
 2015. URL https://arxiv.org/pdf/1409.0473v7.pdf.

Loı̈c Barrault, Ondr̆ej Bojar, Marta R. Costa-jussá, Christian Federmann, Mark Fishel, Yvette Gra-
 ham, Barry Haddow, Matthias Huck, Philipp Koehn, Shervin Malmasi, Christof Monz, Mathias
 Müller, Santanu Pal, Matt Post, and Marcos Zampieri. Findings of the 2019 conference on ma-
 chine translation (wmt19). In Proceedings of the Fourth Conference on Machine Translation
 (Volume 2: Shared Task Papers, Day 1), pp. 1–61, Florence, Italy, August 2019. Association for
 Computational Linguistics. URL http://www.aclweb.org/anthology/W19-5301.

Mauro Cettolo, Jan Niehues, Sebastian Stüker, Luisa Bentivogli, and Marcello Federico. Report on
 the 11th iwslt evaluation campaign, iwslt 2014. In Proceedings of the International Workshop on
 Spoken Language Translation, Hanoi, Vietnam, pp. 57, 2014.

Yen-Chun Chen, Zhe Gan, Yu Cheng, Jingzhou Liu, and Jingjing Liu. Distilling the knowledge of
 bert for text generation. arXiv preprint arXiv:1911.03829, 2019.

Andrew M Dai and Quoc V Le. Semi-supervised sequence learning. In Advances in neural infor-
 mation processing systems, pp. 3079–3087, 2015.

Yuntian Deng, Yoon Kim, Justin Chiu, Demi Guo, and Alexander Rush. Latent alignment and
 variational attention. In Advances in Neural Information Processing Systems, pp. 9712–9724,
 2018.

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep
 bidirectional transformers for language understanding. NAACL, 2019. URL https://arxiv.
 org/pdf/1810.04805.pdf.

Sergey Edunov, Myle Ott, Michael Auli, David Grangier, and Marcaurelio Ranzato. Classical struc-
 tured prediction losses for sequence to sequence learning. NAACL, 2018.

 10
Published as a conference paper at ICLR 2020

Dumitru Erhan, Pierre-Antoine Manzagol, Yoshua Bengio, Samy Bengio, and Pascal Vincent. The
 difficulty of training deep architectures and the effect of unsupervised pre-training. In Artificial
 Intelligence and Statistics, pp. 153–160, 2009.
Dumitru Erhan, Yoshua Bengio, Aaron Courville, Pierre-Antoine Manzagol, Pascal Vincent, and
 Samy Bengio. Why does unsupervised pre-training help deep learning? Journal of Machine
 Learning Research, 11(Feb):625–660, 2010.
Jonas Gehring, Michael Auli, David Grangier, Denis Yarats, and Yann N Dauphin. Convolutional
 sequence to sequence learning. In Proceedings of the 34th International Conference on Machine
 Learning-Volume 70, pp. 1243–1252. JMLR. org, 2017.
Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory. Neural Comput., 9(8):1735–
 1780, November 1997. ISSN 0899-7667. doi: 10.1162/neco.1997.9.8.1735. URL http://dx.
 doi.org/10.1162/neco.1997.9.8.1735.
Marcin Junczys-Dowmunt and Roman Grundkiewicz. Ms-uedin submission to the wmt2018 ape
 shared task: Dual-source transformer for automatic post-editing. EMNLP 2018 THIRD CON-
 FERENCE ON MACHINE TRANSLATION (WMT18), 2018.
Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint
 arXiv:1412.6980, 2014.
Guillaume Lample and Alexis Conneau. Cross-lingual language model pretraining. NeurIPS, 2019.
Guillaume Lample, Myle Ott, Alexis Conneau, Ludovic Denoyer, and Marc’Aurelio Ranzato.
 Phrase-based & neural unsupervised machine translation. arXiv preprint arXiv:1804.07755, 2018.
Gustav Larsson, Michael Maire, and Gregory Shakhnarovich. Fractalnet: Ultra-deep neural net-
 works without residuals. ICLR, 2017. URL https://arxiv.org/pdf/1605.07648.
 pdf.
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike
 Lewis, Luke Zettlemoyer, and Veselin Stoyanov. Roberta: A robustly optimized bert pretraining
 approach. arXiv preprint arXiv:1907.11692, 2019.
Lesly Miculicich, Dhananjay Ram, Nikolaos Pappas, and James Henderson. Document-level neural
 machine translation with hierarchical attention networks. arXiv preprint arXiv:1809.01576, 2018.
Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. Distributed represen-
 tations of words and phrases and their compositionality. In Advances in neural information pro-
 cessing systems, pp. 3111–3119, 2013.
Myle Ott, Sergey Edunov, David Grangier, and Michael Auli. Scaling neural machine translation.
 EMNLP 2018 third conference on machine translation (WMT18), 2018.
Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for automatic
 evaluation of machine translation. In Proceedings of the 40th annual meeting on association for
 computational linguistics, pp. 311–318. Association for Computational Linguistics, 2002.
Jeffrey Pennington, Richard Socher, and Christopher Manning. Glove: Global vectors for word
 representation. In Proceedings of the 2014 conference on empirical methods in natural language
 processing (EMNLP), pp. 1532–1543, 2014.
Matthew E Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and
 Luke Zettlemoyer. Deep contextualized word representations. arXiv preprint arXiv:1802.05365,
 2018.
Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever. Improving language un-
 derstanding by generative pre-training. URL https://s3-us-west-2. amazonaws. com/openai-
 assets/research-covers/languageunsupervised/language understanding paper. pdf, 2018.
Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language
 models are unsupervised multitask learners. OpenAI Blog, 1(8), 2019.

 11
Published as a conference paper at ICLR 2020

Rico Sennrich, Barry Haddow, and Alexandra Birch. Edinburgh neural machine translation systems
 for wmt 16. In Proceedings of the First Conference on Machine Translation, volume 2, pp. 371–
 376, 2016a. URL http://www.statmt.org/wmt16/pdf/W16-2323.pdf.
Rico Sennrich, Barry Haddow, and Alexandra Birch. Improving neural machine translation mod-
 els with monolingual data. ACL, 2016b. URL https://aclweb.org/anthology/
 P16-1009.
Rico Sennrich, Barry Haddow, and Alexandra Birch. Neural machine translation of rare words with
 subword units. ACL, 2016c.
David So, Quoc Le, and Chen Liang. The evolved transformer. In Kamalika Chaudhuri and Ruslan
 Salakhutdinov (eds.), Proceedings of the 36th International Conference on Machine Learning,
 volume 97 of Proceedings of Machine Learning Research, pp. 5877–5886, Long Beach, Cali-
 fornia, USA, 09–15 Jun 2019. PMLR. URL http://proceedings.mlr.press/v97/
 so19a.html.
Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, and Tie-Yan Liu. MASS: Masked sequence to sequence
 pre-training for language generation. In Kamalika Chaudhuri and Ruslan Salakhutdinov (eds.),
 Proceedings of the 36th International Conference on Machine Learning, volume 97 of Proceed-
 ings of Machine Learning Research, pp. 5926–5936, Long Beach, California, USA, 09–15 Jun
 2019. PMLR. URL http://proceedings.mlr.press/v97/song19d.html.
Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov.
 Dropout: a simple way to prevent neural networks from overfitting. The journal of machine
 learning research, 15(1):1929–1958, 2014.
Ilya Sutskever, Oriol Vinyals, and Quoc V Le. Sequence to sequence learning with neural networks.
 In Advances in neural information processing systems, pp. 3104–3112, 2014.
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez,
 Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in neural information
 processing systems, pp. 5998–6008, 2017.
Yiren Wang, Yingce Xia, Tianyu He, Fei Tian, Tao Qin, ChengXiang Zhai, and Tie-Yan Liu. Multi-
 agent dual learning. ICLR, 2019.
Dirk Weissenborn, Douwe Kiela, Jason Weston, and Kyunghyun Cho. Contextualized role interac-
 tion for neural machine translation, 2019. URL https://openreview.net/forum?id=
 ryx3_iAcY7.
Felix Wu, Angela Fan, Alexei Baevski, Yann Dauphin, and Michael Auli. Pay less attention with
 lightweight and dynamic convolutions. In International Conference on Learning Representations,
 2019. URL https://openreview.net/forum?id=SkVhlh09tX.
Lijun Wu, Fei Tian, Yingce Xia, Yang Fan, Tao Qin, Lai Jian-Huang, and Tie-Yan Liu. Learning to
 teach with dynamic loss functions. In Advances in Neural Information Processing Systems, pp.
 6466–6477, 2018.
Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V Le, Mohammad Norouzi, Wolfgang Macherey,
 Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, et al. Google’s neural machine trans-
 lation system: Bridging the gap between human and machine translation. arXiv preprint
 arXiv:1609.08144, 2016.
Yingce Xia, Tianyu He, Xu Tan, Fei Tian, Di He, and Tao Qin. Tied transformers: Neural machine
 translation with shared encoder and decoder. In Proceedings of the AAAI Conference on Artificial
 Intelligence, volume 33, pp. 5466–5473, 2019.
Jiacheng Yang, Mingxuan Wang, Hao Zhou, Chengqi Zhao, Yong Yu, Weinan Zhang, and Lei Li.
 Towards making the most of bert in neural machine translation. arXiv preprint arXiv:1908.05672,
 2019a. URL https://arxiv.org/pdf/1908.05672.pdf.
Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, and Quoc V
 Le. Xlnet: Generalized autoregressive pretraining for language understanding. arXiv preprint
 arXiv:1906.08237, 2019b.

 12
Published as a conference paper at ICLR 2020

A E XPERIMENT S ETUP

A.1 IWSLT’14 & WMT’14 S ETTINGS

We mainly follow the scripts below to preprocess the data: https://github.com/pytorch/
fairseq/tree/master/examples/translation .
Dataset For the low-resource scenario, we choose IWSLT’14 English↔German (En↔De),
English→Spanish (En→Es), IWSLT’17 English→French (En→Fr) and English→Chinese
(En→Zh) translation. There are 160k, 183k, 236k, 235k bilingual sentence pairs for En↔De,
En→Es, En→Fr and En→Zh tasks. Following the common practice (Edunov et al., 2018), for
En↔De, we lowercase all words, split 7k sentence pairs from the training dataset for validation and
concatenate dev2010, dev2012, tst2010, tst2011, tst2012 as the test set. For other tasks, we do not
lowercase the words and use the official validation/test sets of the corresponding years.
For rich-resource scenario, we work on WMT’14 En→De and En→Fr, whose corpus sizes are 4.5M
and 36M respectively. We concatenate newstest2012 and newstest2013 as the validation set and use
newstest2014 as the test set.
We apply BPE (Sennrich et al., 2016c) to split words into sub-units. The numbers of BPE merge
operation for IWSLT tasks, WMT’14 En→De and En→Fr are 10k, 32k and 40k respectively. We
merge the source and target language sentences for all tasks to build the vocabulary except En→Zh.
Model Configuration For IWSLT tasks, we use the transformer iwslt de en setting with
dropout ratio 0.3. In this setting, the embedding dimension, FFN layer dimension and number of
layers are 512, 1024 and 6. For WMT’14 En→De and En→Fr, we use transformer big setting
(short for transformer vaswani wmt en de big) with dropout 0.3 and 0.1 respectively. In
this setting, the aforementioned three parameters are 1024, 4096 and 6 respectively.
Evaluation We use multi-bleu.perl4 to evaluate IWSLT’14 En↔De and WMT translation
tasks for fair comparison with previous work. For the remaining tasks, we use a more advance
implementation of BLEU score, detokenized sacreBLEU for evaluation5 .

A.2 D ETAILED E XPERIMENT S ETTING IN S ECTION 3

The IWSLT’14 English-to-German data and model configuration is introduced in Section A.1.
For the training stategy, we use Adam (Kingma & Ba, 2014) to optimize the network with β1 = 0.9,
β2 = 0.98 and weight-decay = 0.0001. The learning rate scheduler is inverse sqrt, where
warmup-init-lr = 10−7 , warmup-updates = 4000 and max-lr = 0.0005.

A.3 D ETAILED MODEL CONFIGURATION IN U NSUPERVISED NMT

We leverage one Transformer model with GELU activation function to work on translations of two
directions, where each language is associated with a language tag. The embedding dimension, FFN
layer dimension and number of layer are 1024, 4096 and 6. The BERT is initialized by the pre-
trained XLM model provided by (Lample & Conneau, 2019).

B M ORE EXPERIMENT RESULTS

B.1 M ORE RESULTS ON PRELIMINARY EXPLORATION OF LEVERAGING BERT

We use XLM to initialize the model for WMT’14 English→German translation task, whose training
corpus is relative large. We eventually obtain 28.09 after 90 epochs, which is still underperform
the baseline, 29.12 as we got. Similar problem is also reported in https://github.com/
facebookresearch/XLM/issues/32. We leave the improvement of supervised NMT with
XLM as future work.
 4
 https://github.com/moses-smt/mosesdecoder/blob/master/scripts/generic/
multi-bleu.perl
 5
 https://github.com/mjpost/sacreBLEU.

 13
Published as a conference paper at ICLR 2020

B.2 M ORE ABLATION STUDY

Part I: A different way to deal with multiple attention models
Junczys-Dowmunt & Grundkiewicz (2018) proposed a new way to handle multiple attention models.
Instead of using Eqn.(2), the input is processed by self-attention, encoder-decoder attention and
BERT-decoder attention sequentially. Formally,

 ŝlt = attnS (sl−1 l−1 l−1
 t , S
Published as a conference paper at ICLR 2020

the output of each model to get the eventually task. In BERT-fused model, although we include a
pre-trained BERT into our model, there is still only one model serving for the translation task.
In this sense, we can also combine our BERT-fused model with ensemble. Our approach benefits
from ensemble too. When ensembling two models, we can achieve 31.09 BLEU score. When
adding the number of models to four, we eventually achieve 31.85 BLEU score, which is 1.67 point
improvement over the ensemble of standard Transformer.
Part III: More results on IWSLT’14 De→En translation
We report the ensemble results on IWSLT’14 De→En translation in Table 9. We can get similar
conclusion compared to that of IWSLT’14 En→De.

 Table 9: More ablation study on IWSLT’14 De→En.
 Algorithm BLEU
 Standard Transformer 34.67
 BERT-fused model 36.11
 2-model ensemble (standard) 35.92
 3-model ensemble (standard) 36.40
 4-model ensemble (standard) 36.54
 2-model ensemble (BERT-fused) 37.42
 3-model ensemble (BERT-fused) 37.70
 4-model ensemble (BERT-fused) 37.71

B.3 M ORE RESULTS ON FEEDING BERT OUTPUT TO NMT MODULE

The ablation study on more languages is shown in Table 10. Our method achieves the best results
compared to all baselines.

 Table 10: BLEU scores of IWSLT translation tasks.
 Algorithm En→De De→En En→Es En→Zh En→Fr
 Standard Transformer 28.57 34.64 39.0 26.3 35.9
 Feed BERT feature into embedding 29.67 34.90 39.5 28.1 37.3
 Feed BERT feature into all layers of encoder 29.61 34.84 39.9 28.1 37.4
 Our BERT-fused model 30.45 36.11 41.4 28.2 38.7

B.4 M ORE BASELINES OF IWSLT’14 G ERMAN - TO -E NGLISH TRANSLATION

We summarize the BLEU scores on IWSLT’14 De→En of existed works and our BERT-fused model
approach in Table 11.

 Table 11: Previous results of IWSLT’14 De→En.
 Approach BLEU
 Multi-agent dual learning (Wang et al., 2019) 35.56
 Tied-Transformer (Xia et al., 2019) 35.52
 Loss to teach (Wu et al., 2018) 34.80
 Role-interactive layer (Weissenborn et al., 2019) 34.74
 Variational attention (Deng et al., 2018) 33.68
 Our BERT-fused model 36.11

 15
Published as a conference paper at ICLR 2020

B.5 C OMPARISON WITH BACK TRANSLATION

When using unlabeled data to boost machine learning systems, one of the most notable approaches
is back translation (briefly, BT) (Sennrich et al., 2016b): We first train a reversed translation model,
use the obtained model to translate the unlabeled data in the target domain back to source domain,
obtain a synthetic dataset where the source data is back-translated and finally train the forward model
on the augmented dataset.
Our method has two main differences with BT method.

 1. In BT, the monolingual data from the target side is leveraged. In our proposed approach,
 we use a BERT of the source language, which indirectly leverages the monolingual data
 from the source side. In this way, our approach and BT are complementary to each other.
 In Section 5.4, we have already verified that our method can further improve the results of
 standard BT on Romanian-to-English translation.
 2. To use BT, we have to train a reversed translation model and then back translate the mono-
 lingual data, which is time-cost due to the decoding process. In BERT-fused model, we
 only need to download a pre-trained BERT model, incorporate it into our model and con-
 tinue training. Besides, the BERT module is fixed during training.

On IWSLT’14, we also implement BT on wikipedia data, which is a subset of the corpus of training
BERT. The model used for back translation are standard Transformer baselines introduced in Sec-
tion 5, whose BLEU scores are 28.57 and 34.64 respectively. We back translate 1M, 2M, 5M, 15M
and 25M randomly selected German sentences.
The results are reported in Table 12. The rows started with BT(·) represent the results of BT, and the
numbers in the brackets are the number of sentences for back translation.

 Table 12: BLEU scores IWSLT’14 En←De by BT.
 Algorithm En→De
 Standard Transformer 28.57
 BERT-fused model 30.45
 BT (1M) 29.42
 BT (2M) 29.76
 BT (5M) 29.10
 BT (15M) 28.26
 BT (25M) 27.34

IWSLT dataset is a collection of spoken language, and the bilingual training corpus is small (160k).
In Wikipedia, the sentences are relatively formal compared to the spoken language, which is out-
of-domain of spoken languages. We can see that when using 1M or 2M monolingual data for BT,
the BLEU scores can indeed improve from 28.57 to 29.42/29.76. However, simply adding more
wikipedia data for BT does not result in more improvement. There is even a slight drop when
adding more than 15M monolingual sentences. However, our BERT-fused model can achieve better
performances than BT with wikipedia data.

C C OMPARISON OF INFERENCE TIME

 Table 13: Comparisons on inference time (seconds), ‘+’ is the increased ratio of inference time.
 Dataset Transformer Ours (+)
 IWSLT’14 En→De 70 97 38.6%
 IWSLT’14 De→En 69 103 49.3%
 WMT’14 En→De 67 99 47.8%
 WMT’14 En→Fr 89 128 43.8%

 16
Published as a conference paper at ICLR 2020

We compare the inference time of our approach to the baselines. The results are shown in Table 13,
where from the second column to the last column, the numbers are the inference time of standard
Transformer, BERT-fused model, and the increase of inference time.
Indeed, introducing BERT to encode the input brings additional inference time, resulting in about
40% to 49% increase. But considering the significant improvement of BLEU score, it is acceptable
of such extra cost. We will study how to reduce inference time in the future.

D D OWNLOAD LINK OF PRE - TRAINED BERT MODELS
We leverage the pre-trained models provided by PyTorch-Transformers6 .
For IWSLT’14 tasks, we choose BERTbase model with 12 layers and hidden dimension 768.

 1. IWSLT14 En→{De, Es, Fr, Zh}, we choose bert-base-uncased.
 2. IWSLT14 De→En, we choose bert-base-german-cased.

For WMT14 En→{Fr, De}, we choose bert-large-uncased, which is a BERTlarge model with
24 layers and hidden dimension 1024.
For WMT16 Ro→En, we choose bert-base-multilingual-cased, because there is no
BERT specially trained for the Romanian.
For unsupervised En↔Fr and unsupervised En↔Ro, we choose xlm-mlm-enfr1024 and
xlm-mlm-enro1024 respectively.
The download links are summarized as follows:

 • bert-base-uncased: https://s3.amazonaws.com/models.huggingface.co/
 bert/bert-base-uncased.tar.gz.
 • bert-large-uncased: https://s3.amazonaws.com/models.huggingface.
 co/bert/bert-large-uncased.tar.gz.
 • bert-base-multilingual-cased: https://s3.amazonaws.com/models.
 huggingface.co/bert/bert-base-multilingual-cased.tar.gz.
 • bert-base-german-cased: https://int-deepset-models-bert.s3.
 eu-central-1.amazonaws.com/pytorch/bert-base-german-cased.
 tar.gz.
 • xlm-mlm-enfr1024: https://s3.amazonaws.com/models.huggingface.
 co/bert/xlm-mlm-enfr-1024-pytorch_model.bin.
 • xlm-mlm-enro1024: https://s3.amazonaws.com/models.huggingface.
 co/bert/xlm-mlm-enro-1024-pytorch_model.bin.

E D ETAILS OF THE NOTATIONS
Let attn(q, K, V ) denote the attention layer, where q, K and V indicate query, key and value
respectively. Here q is a dq -dimensional vector (d ∈ Z), K and V are two sets with |K| = |V |.
Each ki ∈ K and vi ∈ V are also dk /dv -dimensional (dq , dk and dv can be different) vectors,
i ∈ [|K|]. The attention model works as follows:
 |V |  |K|
 X exp (Wq q)T (Wk ki ) X
 attn(q, K, V ) = αi Wv vi , αi = ,Z= exp((Wq q)T (Wk ki )),
 i=1
 Z i=1
 (6)
where Wq , Wk and Wv are the parameters to be learned. In Vaswani et al. (2017), attn is im-
plemented as a multi-head attention model and we omit the details here to increase readability.
Following Vaswani et al. (2017), we define the non-linear transformation layer as
 FFN(x) = W2 max(W1 x + b1 , 0) + b2 , (7)
 6
 https://github.com/huggingface/pytorch-transformers

 17
Published as a conference paper at ICLR 2020

where x is the input; W1 , W2 , b1 , b2 are the parameters to be learned; max is an element-wise
operator. Layer normalization is also applied following Transformer (Vaswani et al., 2017).

 18
You can also read